Integration of Functional Ex Vivo Drug Testing and Molecular Profiling Reveals Novel Vulnerabilities in CMML

Chronic myelomonocytic leukemia (CMML) denotes a group of rare hematologic neoplasms characterized by peripheral blood monocytosis. Moreover, they have overlapping features with myelodysplastic syndrome (MDS) and myeloproliferative neoplasms (MPN). The genomic landscapes of CMML have been recently described and uncovered few somatic events involving a subset of recurrent gene mutations as TET2, SRSF2, ASXL1, and RAS . Despite the increased knowledge in disease biology, this has not led to deeper understanding of the pathophysiology of the disease mechanisms nor translated in improved outcomes or novel therapeutic modalities. Patients with CMML exhibit significant inter and intra patient heterogeneity with regard to genetic and/or clinical features, which makes identification of disease-specific therapeutic vulnerabilities challenging. Hence, there is need to identify novel treatment modalities and stratify patients according to integrated molecular, functional, and clinical information.

To systematically map the drug sensitivity landscapes and identify novel treatment strategies for CMML, we investigated 14 patient samples derived from 13 CMML patients with a single cell imaging-based ex vivo drug testing approach termed pharmacoscopy (Vladimer et al, Nat Chem Bio 2017). We profiled the sensitivity of 140 clinically relevant, approved and emerging anti-cancer compounds and quantified their effect on total (DAPI), CD14 and CD33 positive cell populations - the later populations representing cancer cells. We scored for CMML cell-selective and differential response by calculating the fraction of marker-positive viable cells in response to drug exposure, utilizing the negative stained population as off-target controls. To further understand and characterize the observed drug effects in a patient-specific manner we performed next generation RNA sequencing as well as targeted gene panel sequencing for myeloid malignancies.

Each CMML sample had a unique drug sensitivity profile. However, we could identify 3 phenotypically distinct subgroups by hierarchical clustering. Several inhibitors were effective in majority of samples such as AT9283 (Aurora A/B; JAK 2/3), volasertib (PLK1), mitoxantrone (Topoisomerase II), and XL228 (IGF1R; SRC; ABL1). Moreover, for certain compounds clustering was observed based on mode of action such as HDAC and PI3K/mTOR inhibitors. Subgroup 1 was characterized by sensitivity to inhibitors targeting HDACs (e.g. romidepsin), CDKs (e.g. flavopiridol), tyrosine kinases (e.g. ponatinib), and farnesyltransferases (e.g. tipifarnib). Subgroup 2 had an overall lower sensitivity to the tested drugs, but displayed vulnerability to the ALK inhibitor crizotinib and DNA and RNA synthesis inhibitor amsacrine. In subgroup 3 volasertib, PI3K inhibitors (e.g. duvelisib), and a subset of DNA synthesis inhibitors (e.g. aminopterin) showed selective efficacy. Interestingly, when comparing our CMML profiling data with in-house drug response data of other hematological malignancies we detected that the top hits are CMML specific. We then investigated differences in the transcriptome between CMML patients and healthy donors and identified profound alterations in gene expression profiles i.e. 1221 genes were significantly upregulated at least 2-fold whereas 718 genes were more than 2-fold downregulated. Upregulated genes were strongly enriched in processes related to cell division and cell cycle progression, while downregulated genes were mainly involved in immune response processes. Moreover, based on transcriptome profiles we could distinguish two clusters of CMML patient samples. Cluster I is composed of CMML patients from subgroup 1 and partially subgroup 2 and is characterized by upregulation of genes involved in TNF signaling. Cluster II includes the remaining samples from subgroup 2 and subgroup 3 defined as upregulation of genes involved in metabolic processes. These findings could be further linked to differential sensitivity to the tested inhibitors.

In summary, we introduce a novel way to stratify CMML patient samples using functional and gene expression profiling, which provides an opportunity to rationally design treatment strategies for individual CMML patients. Deeper integration of this data can serve as a starting point for drug repurposing and generating hypothesis to be investigated in CMML-specific clinical trials.